Hi,
I'm trying to create a dashboard panel that shows the yearly and monthly growth of the cluster
I'm using but it shows non rational numbers
sum(delta(aggr_space_used{}[360d]))
sum(delta(aggr_space_used{}[30d]))
#calculate capacity growth of month and year
ebon portal
wooden apex
Does this chatgpt answer make any sense ? Seems to be not terrible on my setup :
100 *(avg_over_time(aggr_space_used[30d])
/
avg_over_time(aggr_space_used[30d] offset 30d)
- 1
)
Don't shoot the messenger...
Don't assume any of this is correct either, but as always with AI, it seems legit 😄
What delta() Actually Does
delta(v[range])
• Computes last − first value in the range
• Assumes a gauge
• Ignores anything outside the range
• ❌ Does not align “this month vs last month” or “this year vs last year”
So for capacity (absolute disk used), delta() gives:
“How much did disk usage change during this window?”
Not:
“How much bigger is it compared to the same period last month/year?”
❌ Why delta() Is Not Correct for MoM / YoY
Example
If disk usage grows steadily:
Time Capacity
Jan 1 100
Jan 31 130
Feb 28 160
delta(capacity[30d]) → 30
This tells you:
“Disk grew by 30 GB this month”
But MoM growth % should be:
(160 − 130) / 130 = 23%
delta() has no reference to the previous month’s value.
✅ When delta() Can Be Used (Rare)
You can technically compute growth % inside a single window:
100 *
delta(capacity[30d])
/
capacity offset 30d
But this has problems:
Problem Why
Window-bound Breaks if scrape gaps exist
Calendar misalignment “30d” ≠ “previous month”
No YoY support delta() cannot span 1y cleanly
No seasonality Cannot compare same periods
This is not recommended for reporting.
✅ Correct Way (Offset-Based Comparison)
MoM Growth %
100 *
(
avg_over_time(capacity[30d])
/
avg_over_time(capacity[30d] offset 30d)
- 1
)
YoY Growth %
100 *
(
avg_over_time(capacity[30d])
/
avg_over_time(capacity[30d] offset 1y)
- 1
)
Why this is better:
• Aligns periods correctly
• Stable with hourly scrapes
• Works with Grafana time series
• Matches business expectations
🔑 Rule of Thumb
Use case Function
“How much did it grow during this period?” delta()
“How fast is disk growing?” deriv()
“Compare now vs last month/year” offset
“Percent growth MoM / YoY” offset + avg_over_time()
⸻
📌 Final Verdict
No — delta() should not be used for MoM or YoY growth percentages.
It answers a different question.
ebon portal
i’m not looking for mom growth
just simple calc of the capacity growth in TB
Like 100-130
30tb
and the same in yealy
wooden apex
Ah yes not a percentage but an absolute value
ebon portal
yep
wooden apex
Did you try with minus + offset anyways ?
ebon portal
sorry for resolution
its return 0
I can say that some of aggr/node names was change during the last 360d
maybe this causing the issue ?
wooden apex
It might, does your 30d query work ?
ebon portal
I'm not sure if the value is right
i need to check it manually
to be smarter
I think I found something simple that should work sum(aggr_space_used) - sum(aggr_space_used offset 30d)
wooden apex
That's what I would assume yes, you might want to add a grain of avg_over_time for smoother charts (EDIT, you're not working on a chart, so....)
ebon portal
searching by offset in google and found the answer
ebon portal
That's what I would assume yes, you might want to add a grain of avg_over_time f...
tell me what you mean because the graph in the bottom right is not smooth as you can see
it’s just need to show usage in tb over time
wooden apex
I would try to add some avg_over_time() around aggr_space_used offset and aggr_space_used offset 30d
ebon portal
its also catching me the last value of aggregates that got destroyed how can I get just the one that exists?
wooden apex
In dashboards the harvest team use the aggr_new_status in variables, you probably want to filter on this, somehow
Well it seems if you add that variable to your dashboards, and use it as {aggr=~"$Aggregate"} filter it should work
ebon portal
got it